Major haplotype divergence including multiple germin-like protein genes, at the wheat Sr2 adult plant stem rust resistance locus

Plant populations and phenotyping

The generation of the high resolution mapping family between Chinese Spring and the chromosome 3B substitution line Chinese Spring (Hope3B) and the stem and leaf rust rust phenotyping has been described previously [12],[14]. Seeds of emmer (Triticum turgidum var. dicoccum) accessions including cv Yaroslav and of Triticum aestivum cv Marquis and cv Hope were obtained from the Australian Winter Cereal Collection, Tamworth. The Chinese Spring 3BS deletion stock (3BS-8) was obtained from Wheat Genetic and Genomic Resource Center, Kansas State University, USA. The CAP lines were part of the US Wheat Coordinated Agricultural Project (CAP) and kindly provided by Gina Brown-Guedira USDA.

Construction of the 3B-specific BAC library from cv. Hope

The BAC library from chromosome 3B of cv. Hope consisted of two sub-libraries designated TaaHop3BFhA and TaaHop3BFhB, respectively. Both sub-libraries were prepared from the DNA of chromosome 3B purified by flow-cytometric sorting [15],[16] (Additional file 1: Figure S1). Purified genomic DNA was partially digested with HindIII and cloned into the pIndigoBAC-5 vector (Epicentre, Madison, USA). The TaaHop3BFhA sub library was constructed from a limited number of chromosomes (10⁶, corresponding to ~1.8 μg DNA) as described in Šafář et al. [17], with some modifications. The smaller amount of DNA was compensated by lower than normal stringency size selection, including just one size-selection step and usage of DNA fractions of a relatively smaller size (55–80 kb and 80–115 kb). This resulted in a BAC library consisting of 92,160 clones, with an average insert size of 78 kb. This sub-library represented ~6-fold coverage of chromosome 3B. The TaaHop3BFhB sub library was constructed from 2 × 10⁶ chromosomes (~3.6 μg DNA) by applying two size-selection steps as described in Šimková et al. [18]. The higher amount of DNA allowed efficient cloning of larger DNA fragments (145–220 kb), resulting in a library of 43,776 BAC clones with mean insert size of 160 kb, which also represented ~6-fold coverage of chromosome 3B. Screening of the library was done using ³²P labeled DNA probe using standard protocols.

Construction of a non-gridded BAC library from cv. Hope

A second BAC library designated Tae-B-Hope-ng was constructed from cv Hope genomic DNA without chromosome sorting, at the Centre National de Resources Génomiques Végétales (CNRGV)- INRA, France. It was produced using a non-gridded BAC library construction protocol based on Isidore et al. [19] with the following modifications: (1) growth of pooled BAC clones in liquid LB medium; (2) BAC pools DNA amplification by whole genome Genomiphi.v2 phi29 enzyme kit (GE Healthcare) instead of DNA extraction; (3) use of secondary pooling steps in order to identify positive clone coordinates with less screening effort after clone picking. High molecular weight (HMW) DNA was prepared from 20 g of leaf material from cv. Hope. Embedded HMW DNA was partially digested with HindIII (Sigma-Aldrich, St-Louis, Missouri), size selected, eluted and ligated into pIndigoBAC-5 HindIII-Cloning Ready vector (Epicentre Biotechnologies, Madison, Wisconsin). Two successive HMW DNA extractions and three independent sizing steps led to production of 336,670 BAC clones (101 kb average insert size) representing a total ~2-fold coverage of genome. BAC clones were divided into 268 pools before overnight growth and DNA amplification. Pools were screened using specific PCR markers previously developed from clones in the 3B sorted library. The primer sequences for the probe PCR fragments and PCR markers are summarized in Additional file 2: Table S1.

BAC sequencing and sequence assembly

BAC clones from the Sr2 locus of Hope were isolated using probes from the previously characterised 3B locus of Chinese Spring and were sequenced by a BAC-by-BAC shotgun method at 10× coverage using Applied Biosystems capillary sequencing. Base calling, quality trimming and assembly of the raw sequence reads into contigs were done using the PHRED/PHRAP/CONSED package [20],[21] and GAP4 [22] from the Staden package was used for data integration and sequence finishing. The BACs isolated from the non-enriched Hope genomic DNA library were sequenced using Roche 454 sequencing platform. A hybrid assembly of the sequence from 454 and Sanger sequencing was done using Newbler v. 2.3. Additional gap closure was performed on the assembled scaffold sequences by sequencing pooled BAC DNA and an 8 kb paired end library using Roche 454 sequencing platform. Scaffolds were also assembled using Sequencher v. 5.2. Annotated sequence of Sr2 region in Hope has been deposited in GenBank (Acc no. KP244323).

Gene annotation: gene prediction

Genes were annotated using the gene prediction software packages FGENESH (www.softberry.com), AUGUSTUS (http://bioinf.uni-greifswald.de/augustus/) and GENSCAN (http://genes.mit.edu/GENSCAN.html). Gene calls were also made using CLC Genomics workbench v, 7.04. The predicted gene sequences were used for BLASTX and BLASTP analysis against protein databases (NCBI protein and Swiss-prot databases) to identify candidate genes and determine whether they were full length. Intron-exon structure of predicted genes was refined by BLAST comparisons of the predicted gene sequences with the Expressed Sequence Tag (EST) database at NCBI (http://www.ncbi.nlm.nih.gov/dbEST/).

Gene annotation: RNA-seq

Further expressed sequences in the Sr2 region were identified by aligning RNA-seq reads from leaves of cv Hope with the Sr2 genomic sequence. Because of the association of mildew and rust resistance at the Sr2 locus, this RNA was isolated from plants that were inoculated with mildew at four weeks after sowing by applying spores of a glasshouse isolate of Bgt to the abaxial surface of the leaves with a dry paint brush, followed by application of a fine spray of water. Plants were maintained under a shade curtain in the glasshouse at 23°C, in tubs of shallow water to increase humidity, and leaves were sampled at 15 hours after inoculation, and at five days after inoculation. Total RNA was isolated using a Qiagen RNA-Easy Plant Mini-kit, and RNA was DNase-treated on the column using the Qiagen protocol. A pooled sample containing 50% each of DNase-treated RNA isolated from leaves collected at 15 hours and five days after inoculation was sequenced by the Australian Genome Research Facility. RNA-seq was quality trimmed and aligned to the Hope Sr2 genomic sequence using CLC-Genomics Workbench. Expressed regions of the genomic sequence, for which RNA-seq reads aligned at 100% identity, were annotated by BLASTN comparison with DNA sequence databases.

Mapping

Gene sequences identified in the Sr2 region of Hope and CS were mapped as PCR markers (Additional file 2: Table S1) using the high resolution mapping family described previously [12],[14]. The specificity of PCR markers to chromosome 3B was confirmed using the DNA of the Chinese Spring 3BS-8 deletion line. PCRs were performed in a volume of 20 μl with 10 mM of each primer, 0.2 mM dNTPs, 1.5 mM MgCl₂, 1.0 units GoTaq DNA polymerase (Promega) and 100–200 ng of template DNA. The annealing temperature used for PCR was generally 3-4°C below the T_m of the primers used.

A BAC contig for the Sr2region in the stem rust resistant cv Hope

Although the sequence information from the non-Sr2 carrying wheat Chinese Spring (henceforth CS) wheat was useful initially [23],[24], the characterisation of the locus required BAC clones from the Sr2 carrying region of Hope wheat. A BAC library was prepared from flow sorted chromosome 3B of Hope and screened with low copy markers derived from the ends of CS BAC clones. A total of 11 BACs were identified that formed six contigs (Figure 1). Efforts to isolate additional BAC clones from this library, and to close the gaps between the contigs were unsuccessful. To complete the physical map, an additional whole genome BAC library was developed from Hope, and screened with markers flanking the gaps. The positions of the seven newly selected BAC clones were confirmed on the basis of sequence identity with overlapping regions from the BACs derived from the 3B chromosome-sorted library and also genetic mapping of markers which were derived from these new clones. In total, the 18 BAC clones formed a single contig of 968 kb, and included the distal and proximal markers RKO_1 and DOX_1 that recombined with Sr2. The same interval in CS was 567 kb long [24] indicating that the two haplotypes were substantially divergent and included deletion/insertion events. The DNA sequence of the 968 kb Sr2 region (Genbank KP244323) was determined by sequencing all 18 BAC clones.

Gene annotation of the Hope Sr2genomic region sequence and comparison to CS

Based on the level of sequence identity between CS and Hope, the Sr2 locus was divided into two sections (Figures 1 and 2). The proximal section (section 1, 45 kb to 360 kb) of the Hope Sr2 locus sequence showed high similarity with the corresponding region from CS. Eight genes were annotated in section 1 of both Hope and CS, with 2 additional genes present in CS (Table 1) [24]. The length and gene content of this sub-region were closely comparable between CS and Hope, except for a segment of approximately 40 kb in CS, containing a Flowering Locus T– like gene and EMB_2, which were missing in Hope (Table 1). Most of the shared genes in section 1 contained several DNA sequence polymorphisms between the CS and Hope, and their predicted protein products differed by one or more amino acid polymorphisms.

In section two (360 kb to 915 kb; Figures 1 and 2), distal to the shared WD-domain pseudogene, the Hope and CS regions differed markedly. In this region, 26 putative genes were annotated in Hope, whereas only seven genes or gene fragments were annotated in CS. (Table 1). Small regions of high similarity between Hope and CS were evident at 560-590 kb and at 900-968 kb (Figure 1) but apart from these islands of similarity, the two DNA sequences diverged radically. Most notably, an insertion of approximately 100 kb containing ten genes annotated as germin-like proteins (GLPs) was present in Hope, positioned between CS gene ids ctg0011b.000130.1 and 140.1 (Table 1). Despite the lack of overall sequence similarity across the distal part of the locus, other genes that were annotated between WD-domain and DOX_1 were present in both haplotypes (Table 1, Figure 1). This section of the Sr2 locus contained several genes of unknown function (hypothetical proteins, Table 1) that were identified by alignment of the genomic sequence with RNA seq data. These expressed sequences were not annotated as putative genes in the published CS sequence (FN645450), and are listed with the accession number of the most similar GenBank EST in Table 1. For CA696601, CN010964, CA694126, HP626985, CN010964 and BM138011, the sequences in Hope were predicted to encode small proteins (e.g. 84 amino acids for CN010964), but in CS the putative genes were truncated. Overall, the DNA regions that encode these predicted proteins showed 80 to 92% DNA sequence identity between the Sr2 haplotypes. CD882879 encoded a predicted protein of 374 amino acids in Hope and 378 amino acids in CS. Re-sequencing of the gene from CS revealed an error in the deposited FN645450 CS sequence, in which insertion of an extra nucleotide disrupted the open reading frame, resulting in the gene not being annotated in the original CS genomic sequence [24]. The protein coding regions of the CD882879 genes from Hope and CS shared 92% DNA identity, predicting proteins with 90% amino acid identity.

Hypothetical genes D8LAL1 and D8LAL2 were positioned 4 kb from each other in the CS genomic sequence. In contrast, the relative position of these genes was reversed and 60 kb apart in Hope, indicating that inversion and insertion events occurred at this position between the two haplotypes. Predicted genes encoding a fasciclin-like arabinogalactan precursor (FLA; gene id ctg0011b.000180.1), two beta-expansin 1a precursor proteins (BEX_1 and BEX_2; gene ids ctg0011b.000190.1, ctg0011b.000200.1), and a putative disulfide oxidoreductase (DOX_1; gene id ctg0011b.000210.1) formed the distal end of the Sr2 locus in both Hope and CS. All these genes were also present in the corresponding Hope region, as well as an additional copy of the truncated tetratricopeptide repeat domain containing protein, TPR_2 and a beta-expansin precursor, annotated as BEX_3 (Table 1).

Mapping Sr2with respect to newly derived DNA markers

To establish the region that co-segregated with Sr2, PCR markers developed from polymorphic regions of the Hope and CS sequences were mapped in the high resolution mapping family developed by Kota et al. [14]. At the proximal end of the Sr2 region, the marker from the MSF_1 gene (CS gene id ctg0011b.00010.1) was separated from Sr2 in two recombinants, while the marker from the RKO_1 gene (CS gene id ctg0011b.00030.1) was separated by one recombinant. At the distal end, marker RGA_1, derived from a nucleotide binding-leucine rich repeat gene (NB-LRR), was separated from Sr2 in two recombinants, while DOX_1 (CS gene id ctg0011b.000210.1) was separated in a single recombinant. Markers developed from genes between RKO_1 and DOX_1 co-segregated with Sr2. The genes shared between Hope and CS all encoded proteins with one or more polymorphisms between the two haplotypes, and several genes, including the GLPs were specific to Hope. Thus, sequence comparisons alone were unable to pinpoint a specific candidate for Sr2.

Sr2-linked GLPs

The GLP genes co-segregating with Sr2 were of particular interest because of the reported association of GLPs with plant defence ([25] and references therein). Following the convention used for naming GLPs on chromosome 8 of rice [26], the ten GLPs from the Sr2 locus were named TaGLP3_1 to TaGLP3_10 (Triticum aestivum Germin-like proteins from chromosome 3, numbered from proximal to distal). TaGLP3_5, 6 and 10 were pseudogenes because they encoded truncated proteins compared to the other GLPs. Of the remaining seven apparently full-length genes, six, (TaGLP3_ 1/2, −3/4 and - 8/9) were organised as divergently-transcribed pairs, each separated by 610–675 bps between predicted ATG translation start codons, presumably constituting a short, shared promoter (Figure 3). Only TaGLP3_7 was present as a single gene. Sequence comparison showed that the Sr2 locus TaGLP3s could be organised into 2 sequence groups (Table 2 and Figure 4). One group consisted of GLPs 1, 3, 6, 8 and 10 (TaGLP3_1 group), while GLPs 2, 4, 5, 7 and 9 (TaGLP3_2 group) formed the second category. The sequence identity between members within a group ranged between 93-98% while the identity between the GLPs in different groups was 77-82%. For divergently transcribed genes, the gene pair consists of one member of each group. Comparisons of the derived amino acid sequences of the wheat TaGLP3s with sequences from wheat, barley and rice showed that the TaGLP3s were most closely related to a previously undescribed, single, divergently transcribed pair of barley GLPs at a putatively orthologous locus on chromosome 3H (http://plants.ensembl.org/Hordeum_vulgare/Info/Index). Less similar sequences were also present in other cereals, including rice and Brachypodium, although not necessarily at syntenic genomic locations. The TaGLP3s and the orthologous barley GLP gene pair products shared more than 90% amino acid identity with each other, and shared 80 to 84% identity with the next most similar sequences in barley, the Ger4 sub-family, and with the most similar sequences in rice, a group of Ger4 sub-family OsGLP8s on chromosome 8 (Figure 4).

Amplification of TaGLP3s from hexaploid and tetraploid wheat

Sr2 is not related to NB-LRR resistance genes or APR genes Lr34 and Yr36

In this study we have determined the sequence of a region of the wheat chromosome carrying the Sr2 adult plant stem rust resistance gene from cv Hope, and compared DNA sequence and gene content with the non-Sr2 cultivar Chinese Spring that was used as the rust susceptible parent in a high resolution mapping of Sr2 [12],[14]. There were no nucleotide binding site leucine rich-repeat (NB-LRR) genes, the most frequent class of resistance gene, that co-segregated with Sr2, which indicates that Sr2 is not a member of this class. Further, no sequences related to other cloned APR genes such as Lr34 or Yr36 occured in the region [28],[29]. There were many polymorphisms, including large indels between the two haplotypes, preventing identification of Sr2 from sequence data alone.

Haplotype divergence and gene cloning

The Sr2 region of Hope and CS diverged in gene content, including the presence in Hope, and absence in CS, of a cluster of ten GLP genes. Large insertion-deletion events have been reported at other loci in wheat and in other cereals. For example, physical mapping of the adult plant stripe rust resistance locus Yr36 [29] showed a 183 kb insertion in the resistant line RSL65 compared to the susceptible durum wheat Langdon. This region, carrying four genes including the Yr36 resistance gene, originated from the wild species T. turgidum var. dicoccoides. Similarly, a deletion of 351 kb was reported in T. turgidum var. dicoccoides compared to Langdon in the region containing Tsn1, a gene that confers sensitivity to the toxin ToxA in tan spot disease of wheat [30],[31] In rice, large sequence variation between a wild donor species and a recurrent parent was reported for the yield-improving QTL, q-GY2 [32]. In this study, we report major haplotype divergence within the hexaploid wheat species, indicating that large insertion/deletion events may be common, and not restricted to comparisons between inter-species haplotypes. Such major differences will impede rapid, map-based identification of target genes in specific wheat genotypes, even when high quality, reference genome sequence of CS becomes available. It is also likely that haplotype divergence, especially the large insertion/deletion of GLPs in the Sr2 region, resulted in the suppression of recombination and lack of genetic resolution of the target region. Consequently, a new mapping family was established by using Marquis, which carries the GLP gene cluster, as the susceptible parent. Initial results have indicated that in this family, recombination occurred within the region that previously co-segregated with Sr2. The characterisation of new recombinants is in progress.

What is the possible evolutionary value of the presence /absence GLP polymorphism on 3B?

It has been suggested that GLP functional diversity has arisen via sequential gene duplication events, followed by mutation, including in flanking regulatory regions, giving rise to diversity in temporal and spatial gene expression patterns [25]. Diverse wheats reflect four GLP haplotypes in the Sr2 region; Sr2 types with the Hope allele at TaGLP3_9 and TaGLP3_1/2, Marquis types with the Marquis allele at TaGLP3_9 and either the Marquis or Hope allele at TaGLP3_1/2, and CS types with no GLPs at the locus (Tables 2 and 3) [27]. The Hope haplotype with Sr2 and stem rust resistance has been selected by breeders. It is possible that there is an as yet undefined selection acting for or against the Marquis (GLP +) and CS (GLP -) haplotypes in different geographic regions, which may account for the wide spread occurrence of the chromosome 3B presence/absence polymorphism in Sr2- wheat genotypes. This is reminiscent of the presence/absence polymorphism for the resistance gene RPM1 in Arabidopsis thaliana field populations that has been interpreted to result from balancing selection [33].

Germin-like proteins

The GLP genes at the Sr2 locus are of interest because of the reported association of this category of genes with plant defence responses. GLPs are small (~220 amino acid residues), secreted, functionally and taxonomically diverse cupin domain-containing proteins, two of which are reported to produce hydrogen peroxide, a plant defence signal, while several other members of the protein family have been associated with other enzymatic functions [34]. Sequence analysis showed that TaGLP3s at the Sr2 locus were most similar to a pair of divergently-transcribed GLPs at a syntenic position on barley chromosome 3H (Figure 4). The TaGLP3s and this orthologous barley locus, which is distinct from GerD previously mapped on barley chromosome 3H [35], constitute a novel sub-family of GLPs which has not been studied previously, and as far as we are aware, the only sub-family transcribed divergently from short intergenic promoters. The next closest GLP sub-family relative of the Sr2 locus sub-family is the GER4 class of GLP proteins in wheat, barley and rice, members of which have all been functionally associated with fungal disease resistance (Figure 4) ([25] and references therein). Transient over-expression of HvGER4 or HvGER5, as well as transient silencing by RNA interference of HvGER3 or HvGER5 [36] protected barley epidermal cells from attack by powdery mildew. Transient expression of wheat orthologs of HvGER4, germin-gf-2.8, TaGLP2a and TaGLP2b (which have identity between 72-82% with TaGLP3s) in epidermal cells also reduced the penetration efficiency of the powdery mildew fungus Blumeria graminis f. sp. hordei in wheat [37]. No experiments involving over-expression of GLPs in stable transgenic wheat or barley that confirm a direct role of these genes in disease resistance have been reported. Manosalva et al. [26] identified a QTL for rice blast resistance containing a cluster of 12 GLPs on rice chromosome 8, consisting of both GER3 and GER4 sub-family GLPs. RNA interference studies in transgenic rice showed that as more OsGLP genes were suppressed, disease susceptibility of the plants increased. Of the 12 OsGLPs, the OsGLP8-4 member (ortholog of HvGER4) was most effective in providing resistance. The resistance conferred by this QTL was effective against the rice blast and sheath blight pathogens. Despite the sequence similarity between the TaGLP3 gene cluster at the Sr2 locus and GLPs from other cereals that are associated with disease resistance, our data show that the predicted amino acid sequences of the TaGLP3s are identical between the stem rust sensitive cv Marquis, and the resistant cv Hope. Future analysis will determine whether the sequence polymorphisms adjacent to TaGLP3_1,2 or 9 are associated with any differences in gene expression between Hope and Marquis.